1. Field of the Invention
The present invention relates to a digital watermark embedding method, a digital watermark embedding apparatus, a digital watermark detecting method, and a digital watermark detecting apparatus.
2. Description of the Related Art
Popularization of apparatuses for recording and reproducing digital image data such as digital video cassette recorders and DVD (digital versatile disk) systems results in offers of various digital moving picture images which can be reproduced by the apparatus. Furthermore, various digital moving picture images are distributed through digital television broadcasting by using internet, broadcasting satellite, and communication satellite, so that it is being possible for users to use the high quality digital moving picture images.
In the digital moving picture image, since high quality duplicate can be easily produced at a state of a digital signal level, it is a danger of the unlimited duplicate in case that duplicate prohibition or duplicate control is not given. Accordingly, in order to prevent illegal duplicate of the digital moving picture image or in order to control a generation number of the duplicate by normal users, a method for limiting the duplicate, which information for the duplicate control is added to the digital moving picture image and the illegal duplicate is prevented by using the additional information, has been proposed.
The digital watermark is known as a technique which superposes other additional information on the digital moving picture image. In the digital watermarking technique, information (referred to as “watermark information”) such as identifying information of a copyright holder or a user of contents, information of right of the copyright holder, a condition of utilization of the contents, secret information necessary for utilizing, or the above-described information of the duplicate control is embedded on digital contents such as digitalized data of sound, music, moving picture image, and static picture image such that it is not easy to perceive the information. Then, by detecting the watermark information from the embedded contents if necessary, protection of the copyright containing utilizing control and the duplicate control or promotion of secondary utilization can be performed.
Generally, the digital contents express as a set of a plurality of digital data which is a component of the digital contents. For example, in case that the digital contents are a digital image, the digital contents include a set of pixels. The digital contents are expressed as p=(p(z)). Where z of the contents should be a parameter showing a position of the component (for example, pixel), and p(z) should express a data value (for example, pixel value) of the component of the position z.
The digital watermark means other data which are embedded in the contents by varying the data value of one or more components in the contents. As shown generally in the following equation (1), embedding of the digital watermark generates data F[p] having which embedding target contents p is carried out by a transformation F and superposes the data F[p] on the embedding target contents p, which allows embedded contents E[p] to be produced.E[p]=p+F[p]  (1)
On the other hand, detection of the embedded digital watermark generates data G[p′] which detecting target contents p′ is carried out by a transformation G and carries out whether cross-correlation between the data G[p′] and the detecting target contents p′ exceeds a certain threshold value Th1 or not.
                                          p            ′                    *                      G            ⁡                          [                              p                ′                            ]                                      =                              ∑            z                    ⁢                                                    p                ′                            ⁡                              (                z                )                                      ⁢                          G              ⁡                              [                                  p                  ′                                ]                                      ⁢                          (              z              )                                                          (        2        )            
That is,D[p′]=θ(p′*G[p′]−Th1)   (3)where θ(x) is 1 in case of x≧0 (deciding that the digital watermark is in the detecting target contents), and θ(x) is 0 and a step function in case of x<0 (deciding that the digital watermark is not in the detecting target contents), so that a detecting result D[p′] is obtained according to the step function. The transformation G in detecting may be the same as the transformation F in embedding for simplifying.
Scaling is one of typical attacks for erasing or altering the digital watermark (enlargement or reduction in case of the image). When the scaling is carried out, the parameter z assigning the component of the contents is transformed under the scaling of a scaling rate α as follows:z  αz+z0  (4)where, z0 is a constant showing a fixed point for the scaling.
For this reason, the watermark can not be detected even though the transformation G is carried out to find the cross-correlation.
A method called “echo hiding” which is a digital watermarking technique for music is known as a digital watermarking method being robust against the scaling (W. Bender, D. Gruhl, N. Morimoto and A. Lu, “Techniques for data hiding”, IBM Systems Journal, Vol. 35, Nos. 3&4, 313–336, 1996.: hereinafter referred to as “the reference”).F[p](z)=p(z+δ)G[p](z)=p′(z+Δ)  (5)
That is, the parameter z is shifted by δ in case of embedding the digital watermark, and the parameter z is shifted by Δ in case of detecting the digital watermark. In this method, because detection is carried out by varying a value of Δ during detecting the digital watermark, even though the parameter z is transformed by scaling, the detecting result D[p′] of the digital watermark outputs “1” in case of Δ=αδ, namely it is expected that there is a decision that the digital watermark is in the detecting target contents.
However, the reference is not effective in a digital watermark embedding/detecting method in which auto-correlation of the embedded contents has a peak. This is because once an outsider has learned the digital watermark embedding/detecting method, by analyzing whether or not the remarkable (unnatural) peak exists in the auto-correlation, the outsider determines whether or not the digital watermark is in the contents to erase or alter easily the digital watermark. Accordingly, it is desirable to realize a digital watermark embedding/detecting method not having a remarkable peak in the auto-correlation and being robust against the scaling.